Efficient PMV computation for public environments with transient populations

Abstract

Thermal Comfort (TC) is an important environmental parameter strongly affecting human well-being. Nevertheless, it is not being routinely monitored in public environments (e.g., hospitals and shopping malls) characterized by high occupancy and transient populations. Furthermore, the unique computational demands of TC models for such environments are less studied. We establish large datasets representing such settings and corresponding Predicted Mean Vote (PMV) values as calculated by ISO7730 (Fanger’s model). We then demonstrate that PMV values can be reasonably estimated using linear regressions if the full PMV range is piecewise segmented. Support Vector Machine (SVM) regression provides certain accuracy improvement over linear that becomes marginal for sufficiently small segments. However, while SVM computation becomes orders of magnitude slower than ISO7730 algorithm for large datasets, linear computation becomes exponentially faster. Furthermore, the latter does not require unique expertise in mathematics/TC and constitutes an excellent first-step checkpoint to more accurate algorithms adapting the environment to transient populations. Spatial/temporal flexible segment resolution adds compliance with dynamic demands. To conclude, PMV piecewise linear regression can greatly expedite implementing TC and thus conserving energy in public environments, particularly those exposed to extreme climates. To this end, future TC models must consider computation efficiency besides accuracy.